摘要
电力系统的快速发展使得用户对供电可靠性的要求越来越高，配电网故障诊断的重要性尤为突出。若能快速判断故障区段，快速进行故障隔离，及时恢复供电就可以大幅度提高配电网供电的可靠性指标，满足用户对电能质量的要求。配电网故障诊断的研究是当前的研究热点之一。本文将多Agent系统(MAS)这种分布式人工智能方法引入配电网故障诊断研究。由MAS构建实现配电网故障诊断的软件平台，并由有色Petri网(cPN)实现故障区段判断和故障恢复功能。
本文的研究内容和取得的主要成果如下：
1、在深入分析配电管理系统(DMS)、配电自动化系统(DAS)及配电网故障诊断(FD)方法的发展和现状的前提下，指出配电网故障诊断方法需要结合多种方法共同完成的趋势。MAS的分布式求解问题的方法适合解决配电网这种大型网络的分布式问题，同时Petri网的数学表示以及图形表示能够完成对故障诊断过程的描述，因此本文采用基于CPN方法进行故障区域判断，并将结果引入基于MAS的配电网故障诊断平台中进一步得出故障隔离的结果，再交由CPN方法实现故障恢复，最后由MAS平台输出故障恢复策略；
2、在分析配电网拓扑结构特点的基础上，考虑配电网地理信息系统(GIS)在配电网中的广泛应用的现实，提出一种基于GIS的配电网拓扑分析方法；本模型解决商用GIS不能解决配电网拓扑分析的问题，将电力设备与商用GIS的拓扑分析节点对应起来，得出配电网节点描述模型，并通过算例来验证模型的可行性；
3、利用配电GIS的节点描述模型，根据节点的电气属性将节点描述模型转化为元件一开关模型。根据元件一开关模型拓扑分析的特点，提出基于CPN的配电网基本拓扑分析单元。该单元中将配电网转化为元件与开关的连接关系，通过对开关连接情况的描述，针对不同问题得到不同子系统，再对子系统进行分析，得出相应的结论。本文给出实现基于CPN的拓扑分析单元的结构体定义、程序框架及主要函数功能描述；
4、实现配电网故障区段判断即可使用配电网基本CPN拓扑分析单元，此时子系统定义为由不同状态开关连接的元件集合。CPN与配电GIS相关联，数据结构简单，数据内容完整。通过算例对简单故障，复杂故障的故障区段进行判断均取得正确结果。对于含有错误故障信息的开关信息，根据故障后电压的分布情况进行纠错，最终也得到正确结果。在对故障区段进行隔离后，就可以实行故障恢复操作；
5、在分析配电网故障恢复原理的基础上，得出配电网故障恢复过程即为负荷重新分配的过程，因此将基于CPN的基本拓扑分析单元引入配电网故障恢复中。根据启发式规则对配电网进行故障恢复，将启发式规则中的具体实施过程由CPN的基本拓扑分析单元来完成，其中子系统定义为满足故障恢复条件的负荷集合。本算法运算简单，结果清晰明了，在对实际配电网进行仿真实验后，得出正确的结果；
6、提出了基于多Agent的DMS模型，并根据各个Agent模块在交互时对象比较固定，参与交互的Agent较少的情况，采用合同网实现各Agent之间的交互；根据配电网分布式的结构特点，采用MAS构建配电网故障诊断软件平台，给出平台各个Agent模块的功能；详细介绍了故障诊断Agent的实现方法，该Agent采用FIPA体系结构，利用多线程技术实现分块黑板交互模型，并根据实现故障诊断Agent的主体程序框架，设计了主要函数及各个Agent函数的具体功能。
The fast development of the power system causes the user's request to the reliability of the power supply becoming higher and higher, and the fault diagnosis for distribution network is especially important. If we can judge the fault section, then isolate it quickly and restore the power supply promptly, the reliable indexes for power supply will be enhanced greatly and the users' request of the power quality will be ensured. The research on fault diagnosis for distribution network is one of the current research's hot spots. Multi-agent system (MAS) is a distributed artificial intelligence method, and it is used in this paper to research the fault diagnosis of the distribution network. The software platform constructed by MAS is used to complete the fault diagnosis for distribution network and the colored Petri net (CPN) is used to judge the fault section and restore the fault.
The study and the main achievement of this paper as follows:
1．On the base of analyzing the distribution management system (DMS), the distribution automation system (DAS) and the distribution network fault diagnosis (FD) thoroughly, this paper points out that the fault diagnosis for distribution network should be completed by way of synthesizing several methods. The MAS, which uses the distributed method to solve problems, adapts to solve the large-scale network distributed problem, such as distribution network. The mathematical expression and the graphical expression of the Petri net can describe the fault diagnosis process. Therefore the CPN method is used in this paper to judge the fault section. The diagnoses' result is introduced in the distribution network fault diagnosis platform which is based on the MAS, and the further fault isolation result can be obtained. Then the CPN method is used to achieve the fault restoration and the fault restoration strategy is exported through the MAS platform at last;
2．This paper proposes a method of distribution network topology analysis based on geographic information system (GIS) after analyzing the characteristic of distribution network topology structure and considering the reality that GIS is used in the distribution network widely. This model can solve the commercial GIS but it cannot solve the problem of distribution network topology analysis. We can obtain the distribution network node description model by way of corresponding the power equipment's topology analysis nodes and the commercial GIS's. Moreover we can verify the model's feasibility in virtue of calculating an example;
3．By means of the GIS node description model, the GIS node model is transformed into the organ-switch model according to the node organ's electrical attribute. This paper proposes distribution network basic topology analysis unit based on the CPN according to the characteristic of the organ-switch model topology analysis. The distribution network is transformed into the connection between the organ and the switch in this unit. We can obtain the different subsystems in view of the different questions by means of describing the switch's connection situation and then the subsystems are further analyzed to elicit the relevant conclusion. This paper proposes the structure definition of the topology analysis unit based on CPN, the program frame and main function's description;
4．We can use the distribution network's basic CPN topology analysis unit to judge the distribution network fault section and here the subsystem is defined with the combined elements which are connected by different state's switches. Because CPN is connected with GIS, the data's structure is simple and data's content is integrated. We can get the correct conclusion after judging the simple fault and the complex fault section in an example. For the switch information contained the false fault information, we also can obtain the correct conclusion finally after correcting the voltage's distributed situation. After isolating the fault section, we can carry out the fault restoration operation;
5．Based on analyzing the principle of fault restoration, we can obtain the conclusion that the fault restoration's process of distribution network is the load's redistribution process. Therefore the basic topology analysis unit based on the CPN is introduced into the fault restoration of distribution network. The distribution network's fault is restored according to the heuristic rule, and the detailed implement process used heuristic rule relays on the basic topology analysis unit based on CPN. The subsystem is defined with the load combination which satisfies the condition of the fault restoration. This algorithm is simple and the result is clear. We can obtain the correct result after the simulation experiment for the actual distribution network;
6．This paper proposes the DMS model based on MAS, and the contract net is used to realize the alternation between each Agent because the situation that each Agent model is fixed when it alternates and the Agent participating in the alternation is less. According to the distribution feature of the distribution network, this paper uses MAS to construct the distribution network fault diagnosis software platform, and each Agent model's function in the platform is presented in this paper. The realization method of fault diagnoses Agent is introduced in detail. This Agent uses the FIPA system structure and the multi-thread technology to realize block blackboard interactive model. This paper produces the main function and each Agent function's concrete function according to the main body procedure frame using to realize the fault diagnosis Agent.
引文
[1] 刘健．配电自动化系统[M]．北京：中国水利水电出版社，2001，1-4
[2] 王士政．电网调度自动化与配网自动化技术[M]．北京：中国水利水电出版社，2006，228-234
[3] 邵学俭，田磊．基于GIS的专家系统在配电网故障区段判断中的应用[J]．浙江电力，1998，4：12-15
[4] Young-Moon Park, Kwang-Ho Lee, Application of expert system to power system restoration in sub-control Center. IEEE Trans. on Power systems, 1997, 12(2):629-635
[5] 屠强，刘伟等．基于故障投诉信息的配电网故障定位[J]．继电器，2005，33(24)：24-28
[6] 杜红卫，孙雅明等．基于遗传算法的配电网故障区段判断和隔离[J]．电网技术，2000，25(5)：52-55
[7] 孙雅明，吕航．Petri网和冗余纠错技术结合的配网故障区段区段判断新方法[J]．中国电机工程学报，2004，24(10)：61-67
[8] 毕天姝，倪以信等．基于径向基函数神经网络和模糊控制系统的电网故障诊断新方法[J]．中国电机工程学报，2005，25(14)：12-18
[9] 刘淑萍，韩正庆等．基于多Agent的配电网故障处理方案的研究[J]．继电器，2004，32(22)：39-42
[10] 束洪春，孙向飞等．基于粗糙集理论的配电网故障诊断研究[J]．中国电机工程学报，2001，21(10)：73-77
[11] S.Hatakeyama, T.Nagara, H. Sasaki and M.Yasouka, An efficient method for power distribution system restoration based on mathematical programming and operation strategy, Power System Technology, 2000. Proceedings. PowerCon 2000. International Conference on, 2000, 3: 1545 -1550 vol.3
[12] T.Nagata, H. Sasaki and R.Yokoyama. Power system restoration by joint usage of expert system and mathematical programming approach. IEEE Trans. On Power Systems, 1995, 10(3): 1473-1479
[13] 廖志伟，孙雅明等．基于数据挖掘模型的配电网故障区段判断诊断[J]．天津大学学报，2002，35(3)：322-326
[14] Mauro P, Aujan B. A topology that tracks network modifications overtime [J]. IEEE Trans on Power Systems, 1988, 3(3): 992-997
[15] Phongsak D V, Iraj D. A topology-based algorithm for tracking network connectivity [J].IEEE Trans on Power Systems, 1995,10(2): 339-345
[16] 徐晓晖，王超，孙晓龙，梁轩伟，刘荣颖．一种配电网拓扑可观测性理论及其应用[J]．电力电气，2006，25(7)：53-55
[17] 陈竟成，张学松，等．配电网络建模、结线分析及动态网络着色[J]．电力系统自动化，1999，24(23)
[18] 邓佑满，张伯明，汤磊．网络拓扑着色算法在电力系统中的应用[J]．电力系统自动化，1998，22(10)
[19] 陈竟成，徐德超，等．配电网故障恢复系统[J]．电力系统自动化，2000，24(4)：46-51．
[20] 张伟，周步祥．基于有向支路的配电网络拓分析方法[J]．电力系统自动化，2004，28(22)
[21] 刘健，卢建军，蔺丽华．配电自动化不良数据辨识和配电网结线分析[J]．电力系统自动化，2000，24(22)：35-38
[22] 刘健，毕鹏翔，董海鹏．复杂配电网简化分析与优化[M]．北京：中国电力出版社，2002
[23] Liu F, Sun C X, Zhou Q,et al.Study on data model of power network geographic information system[A].Proceedings of the 5st IPEC[C]. Singapore: 2001, 24-28
[24] 宋久旭．基于地理信息系统的配电网络拓扑建模[J]．自动化技术，2005，200(9)：62-64
[25] 廖卫列，刘军，于海玉．基于地理信息系统的配电网络拓扑分析及其应用[J]．电网技术，2006，30(1)：85-88
[26] 杨群，李伟，闾国年．基于GIS构建配电网拓扑关系的方法[J]．电力系统自动化，2003，27(18)：79-82
[27] 孙静，秦世引，宋永华．一种基于Petri网和概率信息的电力系统故障诊断方法[J]．电力系统自动化，2003，27(13)：10-14
[28] 李日隆，李雄刚．利用Petri网对电力系统进行故障诊断[J]．华中电力，2000，13(1)：1-4
[29] 王建元，纪延超，常群等．Petri网络理论在电网故障诊断中的应用[J]．哈尔滨理工大学学报，2002，7(4)：77—80
[30] 赖晓平，周泓兴．电力系统网络拓扑分析的有色Petri网模型[J]．电网技术，2000，24(12)：5-10
[31] 伍艳芳，尉忠信．用于故障诊断的模糊Petri网[J]．航空学报，1995，16(5)：564-568
[32] 陈仕兵，贾再一，龚莲．用有色Petri网模型进行电力系统结线分析[J]．安徽大学学报，2001，25(3)：31-35
[33] 王建元，纪延超．模糊Petri网络知识表示方法在变压器故障诊断中的应用[J]．电力系统自动化，2003，23(1)：123-125
[34] 王雪，杨叔子．基于模糊Petri网络的分布式知识表示方法[J]．华中理工大学学报，1996，24(3)：13-16
[35] 蔡乐，邓佑满，朱小平．改进的配电网故障定位、隔离与恢复算法[J]．电力系统自动化，2001，25(16)
[36] 徐晓晖，孙晓龙等．一种针对数据不全情况的配电网络分析方法及其应用[J]．电气应用，2006，25(4)
[37] 贾丽新．配电网故障恢复与网络重构方法研究：[博士学位本文]．西安：西安交通大学，2003
[38] 束洪春，唐岚，董俊．MAS在电力系统中的应用展望[J]．电网技术，2005，29(6)：27-31
[39] M Wooldridge. Intelligent Agnets: Theory and Practice. Knowledge Engineering Review, 1995, 10(2): 115-152
[40] Y Shoham. Agent-Oriented Programming. Aritificial Intelligence, 1993(50): 51-92
[41] 史忠植．高级人工智能[D]．北京：科学出版社，1998
[42] 田盛丰，黄后宽．人工智能与知识工程[D]．北京：中国铁道出版社，1999
[43] H S Nwana. Software Agent: an overview, Knowledge Engineering Review, 1997, 58(2): 205-245
[44] Foundation for Intelligent Physical Agents. FIPA 97 Speccification. FIPA Architectural Overview. Http://www.fipa.org/specs
[45] 常晋义，陈俊杰，黄磊．组件技术在配电管理系统开发中的应用[J]．计算机工程，2003，29(19)：171-173
[46] Michael Wooldridge，石纯一，张伟，徐晋晖(译)．多Agent引论[M]．北京：电子工业出版社，2003：134-157
[47] 于振华，蔡远利，韩九强．基于面向对象Petri网的MAS交互协议建模[J]．计算机集成制造系统，2005，11(8)：1064-1068